Water resources systems are supposed to serve many different interests, ranging from water supply to conservation of aquatic ecosystems. For this reason, the allocation of water resources is traditionally seen as a multi-criteria problem in which many different interests must be sustainably balanced. The analysis of large-scale water resources systems is often complicated by the presence of multiple reservoirs and flow diversions; the uncertainty of unregulated inflows and demands; and multiple conflicting objectives (e.g., flood control vs. drought prevention vs. protection of the environment) (Haro-Monteagudo et al., 2020). In particular, the uncertainty of inflows makes it impossible to precisely identify future impacts of current decision-making. Hence, the efficient and sustainable operation of multiple reservoir systems is a difficult and challenging task for water resources managers that involves a combination of tactical (management) and strategical (planning) objectives.
Machine learning (ML) approaches have become a common resource in water resources scientists’ toolboxes over the last 25 years, with the ability to process copious quantities of high-dimensional and complex data in a computationally efficient manner being the key attraction of these methods. In recent times, ML has been successfully applied to represent regional hydrological behaviours outperforming traditional hydrological modelling approaches, and even providing insight on non-monitored hydrological process (Kratzert et al. 2019). However, little advances have been made in the consideration of the influence that humans normally exert in water resources systems and the subsequent competition for resources between different water uses, including the environment. In fact, there is still strong resistance in the hydrology community toward adopting these approaches in a serious and fundamental way, with major development groups at international institutions continuing to dedicate most of their efforts to the traditional models that have never benchmarked well against ML (Nearing et al, 2021).
This thesis will contribute to help demystifying the use of ML approaches in the water resources decision-making process by developing approaches that support sustainable planning and operation of large multi-reservoir and multi-purpose systems. The main aim will be to advance towards understanding and solving the complex and longstanding water resource allocation problem. It is expected that this project uses a deep learning (DL), e.g. deep reinforcement learning approach to uncover and interpret relationships in high-dimensional water resources-related data that can enable more accurate streamflow and reservoir storage forecasting to assist in the decision-making processes. This will provide the basis for the future development of DL approaches that incorporate decision-making processes for a sustainable water allocation in complex water resources systems.
Candidates should have (or expect to achieve) a UK honours (or equivalent) at 2.1 or above in Geography, Engineering, Computer Science, Geosciences, Environmental Science, or similar along with excellent written and oral English and basic computer programming skills.
· Basic understanding of concepts in water resources science and/or machine learning methods
· Advanced computer programming skills (Python or R)
Formal applications can be completed online: https://www.abdn.ac.uk/pgap/login.php
• Apply for Degree of Doctor of Philosophy in Geography
• State name of the lead supervisor as the Name of Proposed Supervisor
• State the exact project title on the application form
When applying please ensure all required documents are attached:
• All degree certificates and transcripts (Undergraduate AND Postgraduate MSc-officially translated into English where necessary)
• Detailed CV, Personal Statement/Motivation Letter
Informal inquiries can be made to Dr D Haro Monteagudo (email@example.com) with a copy of your curriculum vitae and cover letter. All general enquiries should be directed to the Postgraduate Research School (firstname.lastname@example.org).
If a suitable candidate is found before the advertised closing date, we reserve the right to withdraw the advert.